Disappearing scientists: Attrition and retention patterns of 2.1 million scientists in 38 OECD countries
Research has been showing that women scientists continue to disappear from science at a significantly higher rate and in higher percentages than men. This is what social scientists have thought for decades鈥攂ut this is no longer the case today, according to a published in Higher Education.
Social scientists still believe in an old and still useful metaphor of the "leaky pipeline," based on numerous interviews and surveys conducted in the U.S. 30 to 40 years ago. There are always more women than men dropping out of education at later stages of their careers鈥攔equiring national and institutional remedies, the traditional narrative goes.
However, structured big data used in the study do not confirm the prevalence of these attrition processes and the power of the leaky pipeline metaphor. What the interview-based U.S. tradition says differs from what the latest analysis of hard data (Big Data) shows
Thirty eight OECD countries were examined, 2.1 million researchers, 11 consecutive cohorts of scientists starting to publish between 2000 and 2010.
A third of scientists in the first cohort (2000) left academia after five years (mainly doctoral school stage) and half of them after 10 years.
Two-thirds of scientists stopped publishing鈥攍eft science in the conceptualization used鈥攚ithin 20 years, with the percentage leaving for all disciplines combined being consistently lower for men. However, behind the aggregate changes at the level of all disciplines, there are nuanced changes at the level of individual disciplines.
The empirical data were provided by Elsevier's ICSR Lab, and Kaplan-Meier survival analysis and associated survival analyses were applied extensively. The studied event was: discontinuation of publication in subsequent years.
In the mathematized STEMM disciplines (MATH, PHYS, COMP, ENG), there is simply no difference between men and women in dropping out of science; the same is true for all STEMM disciplines combined.
The paper provides a good example of longitudinal research in its pure form (rather than repeated cross-sectional studies). This type of study focused on time is only possible with Big Data: tracking the same scientists for up to 22 years.
More information: Marek Kwiek et al, Quantifying attrition in science: a cohort-based, longitudinal study of scientists in 38 OECD countries, Higher Education (2024).
Provided by Adam Mickiewicz University
